Those files are ATMOS encodings of the file you load into REW. REW has everything it needs and does not need to be directly connected, the From file measurement option has all the same data as a direct connection.
Neither is likely to have any effect on your results. The measure with filters feature works as intended. Perhaps you could try connecting a loopback from the output to the input to eliminate any external factors and help isolate where the process is going wrong.
This is the response of the filter you generated:
There's a lot of time delay in your measurements, which you could fix by adjusting the timing offset. I'm not sure what you intended or expected, but when you generate a minimum phase result the original phase of your measurement is discarded...
Which measurement did you select for Measure with these filters? It should be the one used when generating the filters. The resulting measurement will include a list of the filters used and the measurement they were obtained from in the notes.
It should be 12 dB lower (factor of 1/4) to correspond to the increased sample rate and preserve energy. If you import each into REW you should see they have the same frequency response levels.